Integrating fragment-based screening with targeted protein degradation and genetic rescue to explore eIF4E function

Abstract

Eukaryotic initiation factor 4E (eIF4E) serves as a regulatory hub for oncogene-driven protein synthesis and is considered a promising anticancer target. Here we screen a fragment library against eIF4E and identify a ligand-binding site with previously unknown function. Follow-up structure-based design yields a low nM tool compound (4, K_d = 0.09 µM; LE 0.38), which disrupts the eIF4E:eIF4G interaction, inhibits translation in cell lysates, and demonstrates target engagement with eIF4E in intact cells (EC₅₀ = 2 µM). By coupling targeted protein degradation with genetic rescue using eIF4E mutants, we show that disruption of both the canonical eIF4G and non-canonical binding sites is likely required to drive a strong cellular effect. This work highlights the power of fragment-based drug discovery to identify pockets in difficult-to-drug proteins and how this approach can be combined with genetic characterization and degrader technology to probe protein function in complex biological systems.

Fig. 1. Summary of eIF4E protein architecture and protein engineering to identify a suitable system for fragment screening.

a X-ray crystal structure of eIF4E (PDB: 5T46) overlaid with a protein binding partner (eIF4G peptide). Protein secondary structure is represented as coloured ribbons, with eIF4E (yellow), and eIF4G peptide (green). The Connolly surface of m7-GDP is displayed in magenta. b Protein surface of eIF4E showing the canonical binding region as a Connolly surface coloured by hydrophobicity, with red indicating a strongly hydrophobic (lipophilic) region. A fragment of 4E-BP1 is overlaid in ribbon representation (cyan). Residues highlighted (single letter codes) form part of the consensus sequence (YXXXXLφ) for the canonical binding partners, in this case 4E-BP1 (PDB: 3U7X). Where Y = Y54, X = any residue, L = L59, φ = lipophilic residue corresponding to M60 for 4E-BP1. c Protein surface of 4E-BP1-eIF4E fusion protein as a Connolly surface coloured by hydrophobicity.

Fig. 2. Protein-ligand co-crystal structures of compounds 1–4 generated during fragment screening and fragment to lead optimisation.

The secondary structure of eIF4E is shown in ribbon representation. The majority of the protein (grey), canonical peptide derived from 4E-BP1 attached to the N-terminus of eIF4E (blue), loop region between the C-terminus of the α1-helix and N-terminus of the β3-sheet (yellow). The ligand is depicted as green lines. Sidechains from protein residues within 4 Å of the bound ligand are highlighted in orange with single letter codes. Hydrogen bonds between ligand, protein and water are denoted with black dashed lines. Key waters are shown as red spheres and labelled W1 – W4. a Compound 1 bound in site 2 (resolution 1.85 Å). Key features are highlighted including the location of site 2 in comparison to the Cap binding site (the Cap-site ligand (m7-GTP) is shown for illustration purposes only and was not included during protein purification or the screening process). The surface representations of 1 and m7-GTP are shown in cyan and magenta respectively. b Magnified view of binding site 2 with compound 1 bound which has been rotated anti-clockwise by 90°. c Compound 2 structure (resolution 1.89 Å). d Compound 3 structure (resolution 1.93 Å). e Compound 4 structure (resolution 1.97 Å).

Fig. 3. Surface representations of site 2 and the non-canonical binding interface.

Key residues from eIF4E (black) or eIF4G peptide (red italics) are shown as single letter codes. a Compound 4 structure showing the protein Connolly surface coloured by hydrophobicity, red indicating strongly hydrophobic areas. As the site is highly enclosed, the surface associated with W76 and Q81 has been removed to improve visualisation. b Compound 4 structure showing the protein Connolly surface coloured by hydrophobicity and the ligand Connolly surface (yellow). c Overlay of eIF4G peptide (green ribbon) with the protein conformation of compound 4 bound eIF4E (grey ribbon), the reorganised α1-helix and flexible loop region is displayed as a yellow ribbon, the Connolly surface of compound 4 is displayed in cyan. The original loop conformation from the 5T46 structure of eIF4E bound to an eIF4G peptide is displayed as a red ribbon. d Overlay of eIF4G peptide (green ribbon and green Connolly surface) with the protein conformation of compound 4 bound eIF4E. The orientation has been rotated by ~ 90° in the horizontal plane compared to that shown in Figure 3c.

Fig. 4. Compound 4 inhibits eIF4G:eIF4E binding and cap-dependent translation in cell lysate assays.

a Lysates from SW620 cells were incubated with 1–100 µM compound 4 or 5 or positive control peptide (RIIY) for 30 min. Endogenous eIF4E was immunoprecipitated and immunoblotted for eIF4G, 4E-BP1 and eIF4E. Quantitation of 4E-BP1 (b) or eIF4G (c) with endogenous eIF4E in SW620 and HeLa cell lysates, determined by the electro-chemiluminescent binding assay following incubation for 30 min with DMSO vehicle (Cont), 100 µM compound 4 or 100 µM RIIY peptide. Complexes were immobilised by an eIF4E antibody and captured eIF4E, eIF4G and 4E-BP1 detected by their respective secondary antibodies. Values represent ratios of 4E-BP1:eIF4E or eIF4G:eIF4E electro-chemiluminescence relative to DMSO control (n = 2 biological replicates). d Electro-chemiluminescent assay for binding of eIF4G or 4E-BP1 with eIF4E in SW620 (n = 2 biological replicates), or (e) in HeLa lysates (n = 3 biological replicates, mean ± SD) following incubation for 30 min with compound 4 or 5. Results are expressed as luminescence signals relative to DMSO control. f Quantification of eIF4E:eIF4G interaction in H1299 cells by electro-chemiluminescent assay. Cell lysates treated with RIIY 4E-BP1 derived peptide or RIIG negative control peptide at 0.1–100 µM for 30 min (n = 2 biological replicates). g Quantification of the endogenous eIF4E:eIF4G interaction in H1299 cell lysates at 0.1–100 µM (for 6 h) of compound 4 or 5, as measured by electro-chemiluminescent assay (mean ± SD from n = 3 biological replicates). h HeLa cell lysates for in vitro translation were incubated for 30 min with 1, 10, 100 µM of compound 4 or 5. Results are expressed as firefly or renilla luminescence signal normalized to DMSO control and expressed as % (mean ± SD from n = 3 biological replicates). Significance was determined using two-sided unpaired t-test comparing compound 4 to compound 5 at each concentration. Statistically significant p-values (p < 0.05) are shown on the plot and source data is located in the Source Data file.

Fig. 5. Compound 4 binds eIF4E in intact H1299 cells.

a Top: Representative immunoblot (n = 2 biological repeats) showing eIF4E protein stabilisation in H1299 cells at 57.6 °C following compound treatment. Cells were treated with compound 4 or compound 5 (50 μM) for 6 hrs and then incubated at 57.6 °C. Bottom: Representative immunoblot (n = 3 biological repeats) following treatment with compound 4 (10, 5, 1, 0.5 μM) or compound 5 (10 μM) for 6 hrs followed by incubation at 57.6 °C. Vinculin was used for loading control. b eIF4E stabilisation in H1299 cells at 57.6 °C following compound 4 treatment was quantified from immunoblots using Image J and normalised to vinculin control. Statistically significant p values (p < 0.05) are shown on the plot (mean ± SEM from n = 3 biological replicates). c Quantification of the endogenous eIF4E:eIF4G interaction in intact H1299 cells treated at 100 µM for 6 or 16 hr (mean ± SD from n = 3 biological replicates). Statistically significant p-values (p < 0.05) are shown on the plot. d Cells were transfected with a protein synthesis reporter expressing a bicistonic mRNA with a cap-dependent luciferase reporter (fLuc) and a cap-independent luciferase (rLuc) driven by a viral IRES and were treated with compound 4, 5 (10 or 25 µM) or cycloheximide (CHX; 100 µM) for 24 hr. Luminescence was measured using the Dual-glo luciferase assay. All p values for compounds 4 and 5 were not statistically significant (p > 0.05; mean ± SEM from n = 3 biological replicates). e Cell viability measured by cell titre blue assay in cells treated with compound 4 or 5 for 4 days (mean ± SEM from n = 3 biological replicates). For (b–d) significance (p < 0.05) was determined using an ordinary one-way ANOVA with Tukey multiple comparisons test, comparing compound 4 treatment with the control or negative control compound 5 treatment conditions. Source data is located in the Source Data file.

Fig. 6. Mutants of site 2 disrupt binding of eIF4G in H1299 cells.

a Crystal structure of eIF4E bound to a 35-residue fragment of eIF4G (PDB: 5T46) showing key features previously described in Figs. 1a and 2a. Highlighted in cyan are the residues selected for mutational analysis which gave suitable expression levels (W56, W73, L85, L134 and S209). b Quantification of eIF4E:eIF4G interaction determined by the electro-chemiluminescent assay in cells transfected with wild type (WT) eIF4E or a series of eIF4E mutants. c Quantification of N-terminal FLAG-tagged eIF4E expression in cells transfected with wild type (WT) eIF4E or a series of eIF4E mutants. d Quantification of eIF4G normalised to eIF4E (data from b and c) in cells transfected with wild type (WT) eIF4E or a series of eIF4E mutants. All plots (b–d) show mean ± SD from n = 3 biological replicates with significance for each mutant relative to the WT control using an ordinary one-way ANOVA with Tukey multiple comparisons test. Significant p-values (p < 0.05) are shown on the plots and the source data is located in the Source Data file.

Fig. 7. Characterization of an eIF4E dTAG-degradation model.

a Schematic of the generation of stable dTAG eIF4E single clones in H1299 cells for eIF4E degradation (created in BioRender. Powers, M. (2024) https://BioRender.com/t36s392). A stable cell pool expressing eIF4E-dTAG was established, endogenous eIF4E was removed using CRISPR/Cas9, and single cell clones were isolated. Treatment of isolated clones with the heterobifunctional dTAG^V−1 molecule recruits E3 ligase VHL to degrade eIF4E-dTAG. Four eIF4E-dTAG CRISPR clones were selected: two C-terminal FLAG-tagged (C2-2, C3−1) and two N-terminal FLAG-tagged (N2−1, N3-2). b Immunoblot analysis of eIF4E, FLAG-tagged eIF4E, and MCL1 expression. Parental H1299 cells and the selected eIF4E-dTAG clones were treated with 0.5 and 1 µM dTAG^V−1 for 6, 8, or 16 h. Vinculin was used as a loading control (n = 3 biological replicates). c Real-time cell growth measurements of the selected eIF4E-dTAG clones treated with 0.5 or 1 µM dTAG^V−1. Cell confluency (%) was monitored every 4 h over 6 days using Incucyte Zoom. Significance was determined using a one-way ANOVA with Tukey multiple comparisons test comparing dTAG^V−1 treatment with the control in each cell line. Significant p values (p < 0.05) are shown (mean ± SEM from n = 3 biological replicates), the source data is located in the Source Data file. d Parent H1299 cells were treated with eIF4E siRNA (1 µg; eIF4E siRNA) or control siTOOLs (PAR) for 72 h (n = 3 biological replicates). The four eIF4E-dTAG clones were treated with vehicle (Con no dTAG) or 500 nM dTAG^V−1 (dTAG) for 72 h (n = 1 repeat from each individual clone) and global proteomes profiled. Significant differences between control and treated samples were determined using MSstats. Plots show signal intensity data for eIF4E (CON no dTAG v dTAG and eIF4E siRNA v PAR both p_adj < 0.05) and MCL1 (p_adj> 0.05; not significant). e Log₂[signal intensity] for proteome profiles of the individual eIF4E-dTAG clones with the control parent H1299 cells. Each point represents the ratio of individual protein expression between the control parent and different control dTAG clones (C2-2, C3-1, N2−1 and N3-3). Diagonal lines represent a 1.5-fold change in protein expression (Supplementary Data 5).

Fig. 8. Genetic rescue of eIF4E dTAG degradation through expression of wild type eIF4E or mutant of the different functional domains of eIF4E.

eIF4E dTAG C3-1 clone or C3−1 were transfected with wild-type eIF4E (WT) or a series of eIF4E mutants and treated with 0.5 or 1 µM dTAG^V-1. a Quantification of MCL1 immunoblot signal using Image J. Results are expressed as % MCL1 normalized to vinculin (mean ± SEM from n = 3 biological replicates) following 6 hr dTAG^V−1 treatment (Supplementary Fig. 17b). One sample t and Wilcoxon tests (two-sided), using 100% as hypothetical control value, were performed and compared with the control in each cell line. b, c Real-time cell growth measurements monitored every 4 h over a 6-day period. Cell confluency (%) was calculated using Incucyte Zoom software and (mean ± SEM n = 3 biological replicates). Ordinary one-way ANOVA with Tukey multiple comparisons test compared treatment with the control in each cell line. Significant p-values (< 0.05) are shown on the plots. d Representative immunoblot of 3 biological replicates of eIF4E and MCL1 protein expression following 16 hr treatment of dTAG^V-1 (1 µM) or compound 4 (25 μM) or in combination in H1299 parental cells, the eIF4E dTAG C3-1 clone or C3-1 transfected with W73F or L85R mutant. Vinculin was used as loading control. H1299 (PAR) cell lines were treated with 1 μM dTAG^V-1 only as qualitative control. Vinculin was used as loading control. Quantification from MCL1 immunoblots normalised to control is shown in Supplementary Fig. 19c. e Real-time cell growth of the eIF4E dTAG C3-1 clone and C3-1 transfected with W73F or L85R mutant following treatment with 25 µM compound 4 ± 1 µM dTAG^V-1 were monitored and analysed as described for (c). Significant p-values (< 0.05) are shown on the plots and the source data is located in the Source Data file.